Typical chess engines calculate the lines assuming the highest strength for both sides. I need a chess engine, which calculates the lines, while assuming one of the sides is weaker. For example, it assumes White is 3000 and Black is 1600, and calculates lines with that assumption. All of its lines would probably give something like +4 advantage to White. It will be useful, since it will show the mistakes that Black could make in a position, and the ways White could exploit them. Is there any chess engine that allows something like this?
When I look at a position I decide whether it's drawn or winning without looking at my opponent's strength. You do it differently? You don't know whether a position is good or bad unless you can see your opponent's rating? And you think computers are the same way? Because I don't think a computer engine makes assumptions about strength.
I don't think that's a concept that's even defined. If a computer made an assumption of strength 1600 what specifically would it do differently?
More importantly why do you think making an assumption of strength, whatever you think that means, would accomplish something worthwhile for you?
I don't think that's a concept that's even defined. If a computer made an assumption of strength 1600 what specifically would it do differently?
More importantly why do you think making an assumption of strength, whatever you think that means, would accomplish something worthwhile for you?
@StingerPuzzles said in #2:
> When I look at a position I decide whether it's drawn or winning without looking at my opponent's strength. You do it differently? You don't know whether a position is good or bad unless you can see your opponent's rating? And you think computers are the same way? Because I don't think a computer engine makes assumptions about strength.
>
> I don't think that's a concept that's even defined. If a computer made an assumption of strength 1600 what specifically would it do differently?
>
> More importantly why do you think making an assumption of strength, whatever you think that means, would accomplish something worthwhile for you?
Suppose this is the position you want to analyze:
lichess.org/analysis/r1bq1rk1/ppp2ppp/2np1n2/2b1p1B1/2B1P3/3P1N2/PPP2PPP/RN1Q1RK1_w_Qq_-_0_1#1
The engine evaluates the position as +0.3 and shows a somewhat forcing line as the top line. That teaches me very little about the plans in the position. If I want to learn how to play the position as White, I should play some moves that I find logical for Black, and ask the engine to show me how should White respond to them and exploit the potential mistakes.
I want the engine do this for me. That is, instead of showing the best continuation, I want the engine to show 7 or 8 possible continuations that would happen if someone of my strength were to play against the engine. That will help me to find the plans and traps in the position, and help me prepare to exploit the mistakes that my opponent will possible make against me in the position.
> When I look at a position I decide whether it's drawn or winning without looking at my opponent's strength. You do it differently? You don't know whether a position is good or bad unless you can see your opponent's rating? And you think computers are the same way? Because I don't think a computer engine makes assumptions about strength.
>
> I don't think that's a concept that's even defined. If a computer made an assumption of strength 1600 what specifically would it do differently?
>
> More importantly why do you think making an assumption of strength, whatever you think that means, would accomplish something worthwhile for you?
Suppose this is the position you want to analyze:
lichess.org/analysis/r1bq1rk1/ppp2ppp/2np1n2/2b1p1B1/2B1P3/3P1N2/PPP2PPP/RN1Q1RK1_w_Qq_-_0_1#1
The engine evaluates the position as +0.3 and shows a somewhat forcing line as the top line. That teaches me very little about the plans in the position. If I want to learn how to play the position as White, I should play some moves that I find logical for Black, and ask the engine to show me how should White respond to them and exploit the potential mistakes.
I want the engine do this for me. That is, instead of showing the best continuation, I want the engine to show 7 or 8 possible continuations that would happen if someone of my strength were to play against the engine. That will help me to find the plans and traps in the position, and help me prepare to exploit the mistakes that my opponent will possible make against me in the position.
You are looking for the best move on the board, independent of the strength. Stronger players only have let us say 10k more games than a beginner on his belt. He studies chess, analyses his games and so on. And different players look at the position differently. One can play more positional another more tactical.
You can let the engine show you more than one move, as well you can play the move you think is right, and see what the answer for this move is, and you can play for both sides as well and let the engine show you whats wrong with the moves. It does not make any sense to have a stronger side than the another. Or you can ask a colleague of yours, to go through your game and see what he would play at such position.
You can let the engine show you more than one move, as well you can play the move you think is right, and see what the answer for this move is, and you can play for both sides as well and let the engine show you whats wrong with the moves. It does not make any sense to have a stronger side than the another. Or you can ask a colleague of yours, to go through your game and see what he would play at such position.
At the position what you showed it is the giouco piano/italian opening. You can open the opening explorer to take a look in the main lines of the opening. Here the position is almost symmetrical.
In my humble opinion:
White:
- almost complete dev, pinning the kings knight.
- plans:
— complete dev, bring some tactics in play (nc3-d5), break the castle (trade bishop for the knight when the knight is only defended by the pawn), king side attack.
- black
- almost complete dev, with the kings knight pinned
- plans:
— complete dev, break the pin without weakening the castle, defend the knight with the rook (be6 offering bishop trade, getting the f-file open), re-route the queens knight to defend the kings knight, attack the center or the queen side, do not allow whites plan
These are some of my plans, take it with a little bit of salt. You can take a look in the Wikipedia/eco/mco... as well to see what kind of plan it is recommended.
Thx
In my humble opinion:
White:
- almost complete dev, pinning the kings knight.
- plans:
— complete dev, bring some tactics in play (nc3-d5), break the castle (trade bishop for the knight when the knight is only defended by the pawn), king side attack.
- black
- almost complete dev, with the kings knight pinned
- plans:
— complete dev, break the pin without weakening the castle, defend the knight with the rook (be6 offering bishop trade, getting the f-file open), re-route the queens knight to defend the kings knight, attack the center or the queen side, do not allow whites plan
These are some of my plans, take it with a little bit of salt. You can take a look in the Wikipedia/eco/mco... as well to see what kind of plan it is recommended.
Thx
So you want the computer to do what you would do. Now it goes away from an engine and becomes an AI. Because the engine will ask "What is a rating?". Now the AI has to determine what your rating(and probably playing style to get a "personalised" line)is, then process what the engine will take as input to show you the line you want, and then pass it to the engine. Seems like a lot of work.
@magicsacrifblunder said in #5:
> At the position what you showed it is the giouco piano/italian opening. You can open the opening explorer to take a look in the main lines of the opening. Here the position is almost symmetrical.
> In my humble opinion:
> White:
> - almost complete dev, pinning the kings knight.
> - plans:
> — complete dev, bring some tactics in play (nc3-d5), break the castle (trade bishop for the knight when the knight is only defended by the pawn), king side attack.
> - black
> - almost complete dev, with the kings knight pinned
> - plans:
> — complete dev, break the pin without weakening the castle, defend the knight with the rook (be6 offering bishop trade, getting the f-file open), re-route the queens knight to defend the kings knight, attack the center or the queen side, do not allow whites plan
>
> These are some of my plans, take it with a little bit of salt. You can take a look in the Wikipedia/eco/mco... as well to see what kind of plan it is recommended.
>
> Thx
Thanks for the response.
Let's suppose White plays Nc3 in the position (lichess.org/analysis/r1bq1rk1/ppp2ppp/2np1n2/2b1p1B1/2B1P3/3P1N2/PPP2PPP/RN1Q1RK1_w_Qq_-_0_1#0)
Now I want the engine to make a list like this:
- After Qd8 by black, Bxf6 gives White +2.0 advantage.
- After g6 by black, Nd5 gives White +7.5 advantage.
- ...
That is, I want the engine show me the ways my opponent (with a certain strength) might make mistakes. Normally, I can achieve this by trying different moves and checking the engine's response, but I want to make engine do it for me.
> At the position what you showed it is the giouco piano/italian opening. You can open the opening explorer to take a look in the main lines of the opening. Here the position is almost symmetrical.
> In my humble opinion:
> White:
> - almost complete dev, pinning the kings knight.
> - plans:
> — complete dev, bring some tactics in play (nc3-d5), break the castle (trade bishop for the knight when the knight is only defended by the pawn), king side attack.
> - black
> - almost complete dev, with the kings knight pinned
> - plans:
> — complete dev, break the pin without weakening the castle, defend the knight with the rook (be6 offering bishop trade, getting the f-file open), re-route the queens knight to defend the kings knight, attack the center or the queen side, do not allow whites plan
>
> These are some of my plans, take it with a little bit of salt. You can take a look in the Wikipedia/eco/mco... as well to see what kind of plan it is recommended.
>
> Thx
Thanks for the response.
Let's suppose White plays Nc3 in the position (lichess.org/analysis/r1bq1rk1/ppp2ppp/2np1n2/2b1p1B1/2B1P3/3P1N2/PPP2PPP/RN1Q1RK1_w_Qq_-_0_1#0)
Now I want the engine to make a list like this:
- After Qd8 by black, Bxf6 gives White +2.0 advantage.
- After g6 by black, Nd5 gives White +7.5 advantage.
- ...
That is, I want the engine show me the ways my opponent (with a certain strength) might make mistakes. Normally, I can achieve this by trying different moves and checking the engine's response, but I want to make engine do it for me.
@vahid96 so you want the analyse tree of all the moves. You know that the number of possibilities of moves increases exponentially with the depth. Right? You will get a bunch of information that you will discard anyway. The best is to learn the best line, and this the engine can give. Or like i said, try the lines that you think that it is good, and see what the engine tells you.
Or make an engine for yourself, with the list of all possibilities. I am not sure if sf has an interface to list all the moves.
Or make an engine for yourself, with the list of all possibilities. I am not sure if sf has an interface to list all the moves.
@BravestKnight said in #6:
> So you want the computer to do what you would do. Now it goes away from an engine and becomes an AI. Because the engine will ask "What is a rating?". Now the AI has to determine what your rating(and probably playing style to get a "personalised" line)is, then process what the engine will take as input to show you the line you want, and then pass it to the engine. Seems like a lot of work.
There are already engines with lower strengths, I don't think it is necessary for it to be an AI.
Consider this position (that I have already sent many times): lichess.org/analysis/r1bq1rk1/ppp2ppp/2np1n2/2b1p1B1/2B1P3/3P1N2/PPP2PPP/RN1Q1RK1_w_Qq_-_0_1#1
The engine evaluates h6 by Black as +0.2. But if Black follows it with g5, then White should sacrifice the Knight and White will have a +1.1 advantage. So when the engine evaluates h6 as +0.2, the fact that Black could easily make a mistake afterwards remains hidden to the human.
In a way, I need the engine to show me all the lines that lead to a +2 advantage for white after some moves, but the moves by Black shouldn't be obviously bad moves.
> So you want the computer to do what you would do. Now it goes away from an engine and becomes an AI. Because the engine will ask "What is a rating?". Now the AI has to determine what your rating(and probably playing style to get a "personalised" line)is, then process what the engine will take as input to show you the line you want, and then pass it to the engine. Seems like a lot of work.
There are already engines with lower strengths, I don't think it is necessary for it to be an AI.
Consider this position (that I have already sent many times): lichess.org/analysis/r1bq1rk1/ppp2ppp/2np1n2/2b1p1B1/2B1P3/3P1N2/PPP2PPP/RN1Q1RK1_w_Qq_-_0_1#1
The engine evaluates h6 by Black as +0.2. But if Black follows it with g5, then White should sacrifice the Knight and White will have a +1.1 advantage. So when the engine evaluates h6 as +0.2, the fact that Black could easily make a mistake afterwards remains hidden to the human.
In a way, I need the engine to show me all the lines that lead to a +2 advantage for white after some moves, but the moves by Black shouldn't be obviously bad moves.
@magicsacrifblunder said in #8:
> @vahid96 so you want the analyse tree of all the moves. You know that the number of possibilities of moves increases exponentially with the depth. Right? You will get a bunch of information that you will discard anyway. The best is to learn the best line, and this the engine can give. Or like i said, try the lines that you think that it is good, and see what the engine tells you.
>
> Or make an engine for yourself, with the list of all possibilities. I am not sure if sf has an interface to list all the moves.
I think there might be different ways to achieve that. An example would be to command the engine to always follow the best line for White and follow the third best line for Black.
I suppose engines calculate many lines, but they don't show the discarded lines. Maybe if an engine shows these discarded lines, it can be used for my purpose (There will be many discarded lines, but there may be ways to find the more interesting ones and show them only).
> @vahid96 so you want the analyse tree of all the moves. You know that the number of possibilities of moves increases exponentially with the depth. Right? You will get a bunch of information that you will discard anyway. The best is to learn the best line, and this the engine can give. Or like i said, try the lines that you think that it is good, and see what the engine tells you.
>
> Or make an engine for yourself, with the list of all possibilities. I am not sure if sf has an interface to list all the moves.
I think there might be different ways to achieve that. An example would be to command the engine to always follow the best line for White and follow the third best line for Black.
I suppose engines calculate many lines, but they don't show the discarded lines. Maybe if an engine shows these discarded lines, it can be used for my purpose (There will be many discarded lines, but there may be ways to find the more interesting ones and show them only).
This topic has been archived and can no longer be replied to.